Land use/cover classification of arid and semi-arid Mediterranean landscapes using Landsat ETM

Land use/cover classification in heterogeneous east Mediterranean landscapes is challenging, e. g. Jordan. Digital land use/cover maps are needed at an appropriate cost, spatial and temporal coverage. North-western Jordan is appropriate for exploring the use of Landsat ETM imagery in land use/cover delineations as its biogeographic zones are very diverse and heterogeneous. Supervised and unsupervised classification schemes were used with and without spatial enhancement techniques. Sources of classification errors were inspected statistically. Results indicated that Landsat ETM images are effective in classifying heterogeneous Mediterranean landscapes with an accuracy of up to 83%. Accuracy was enhanced by approximately 9% using supervised classification. Spatial enhancement improved accuracy of certain classes and reduced it for others. Results call for class-specific classification schemes. Areas of the different land use/cover classes of the study area were estimated from the classified image. Urban, shrubland and rangeland areas were estimated for the first time. There is a growing concern about the governmental census estimates of certain classes.